See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/288640223

Considerations for Hydrodynamic Slug Analysis in Pipelines

Article · September 2014

DOI: 10.1115/IPC201433098

CITATIONS READS

0 823

5 authors, including:

Daniela Gálatro Xiangyu Hu

13 PUBLICATIONS   19 CITATIONS   
Technische Universität München
180 PUBLICATIONS   3,715 CITATIONS   
SEE PROFILE
SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A high-end multiphysyics smoothed particle hydrodynamics View project

level-set method View project

All content following this page was uploaded by Xiangyu Hu on 10 May 2017.

The user has requested enhancement of the downloaded file.

 DRAFT
International Pipeline Conference
IPC2014
September 29- October 3, 2014, Calgary, Alberta, Canada

IPC2014-33098

CONSIDERATIONS FOR HYDRODYNAMIC SLUG ANALYSIS IN PIPELINES

Daniela Galatro Daniel Fürtbauer

ILF Beratende Ingenieure ILF Beratende Ingenieure
Department of Process Simulation Department of Process Simulation
Munich, Bavaria, Germany Munich, Bavaria, Germany
Email: [email protected] Email: [email protected]

Xiangyu Hu David-Emilio Nerucci

Technische Universität München ILF Beratende Ingenieure
Institute of Aerodynamics Department of Pipeline Technology and System
Munich, Bavaria, Germany Design
Email: [email protected] Munich, Bavaria, Germany
Email: [email protected]

Flavio Marín
ILF Beratende Ingenieure
Department of Pipeline
Technology and System Design
Munich, Bavaria, Germany
Email: [email protected]

ABSTRACT the pipeline profile. The results yielded by this model have been
Hydrodynamic slugs in pipelines are usually analyzed by compared with field data and results performed by using a
using a steady-state flow assurance simulator as a first transient simulation software, showing fairly accurate values.
approximation. The pipelines are then modeled in transient
simulation software to get more accurate values. Comparisons INTRODUCTION
between empirical and mechanistic method are made in this Slug occurs in many engineering applications such as
work by running simulations in the steady-state simulators in transport of hydrocarbon fluids in pipelines or partial
order to explain the differences in the calculated slug properties. vaporization of fluids. The slug flow formation is strictly of
Additionally, slug sizes for operational slugging have been transient nature [1, 2].
analyzed by using a new alternative pseudo transient approach Slug flow represents a risk to the structure of a pipe system
to the Lagrangian slug tracking scheme. The model expresses and might lead to severe or undesired issues like large liquid
an unsteady state mass balance in a pipeline, formulated phases, flow rates larger than the average value, considerable
utilizing the slip velocity written in terms of the void fraction momentum over pipe bends due to long slugs, large pressure
and superficial gas velocity. Our model includes a constitutive drops in a section of the pipeline and resonance problems [6].
equation for slip velocity, elevation changes to represent the Hydrodynamic slugs are initiated by the instability of
hydraulic profile of the pipeline, a method for the calculation of waves on the gas-liquid interface in stratified flow at certain
the maximum slug length, a modified correlation for the slug flowing conditions [3]. As gas passes over a wave there is a
length calculation and the variation of the fluid density along pressure drop; a small force upward within the wave is created

1 Copyright © 2013 by ASME

due to the pressure recovery. Under certain flowing conditions, H Lslug E
this force lifts the wave until it reaches the top of the pipe. Lbubble = ∫H Lfilm F
dH L (1)
Two different approaches have been developed and used to
analyze hydrodynamic slugs in pipelines: steady-state and
transient. Where: E and F are functions of the liquid and gas
Numerous empirical models for predicting multiphase flow densities, cross-sectional areas, shear stresses and pipe
properties have been developed. A comprehensive review of inclination.
these models is given by Brill and Mukherjee [9]. Depending on The average film holdup is given by:
the complexity of the empirical model it can contain equations
for determining the flow regime and predicting the liquid _____ H Lslug E
holdup and pressure gradient in the predominant flow pattern. Lbubble H Lfilm = ∫H Lfilm F
H L dH L (2)
Slug properties such as slug length and slug frequency can also
be determined with empirical correlations.
On the other hand, computational codes of transient two- The following settings are applied in the steady-state
phase flow have been developed since the early nineties; one of commercial software used in this work:
the most extensively used codes is OLGA [10]. These codes are
based on the two-fluid models (conservation equations) or the Slug Holdup. The slug holdup is obtained from the
drift flux model. correlation presented by Gregory et. al [12].
Transient events in multiphase flow models such as ramp-
up and pigging have been fairly modelled by using transient and Slug Translational Velocity. The slug translational
steady-state approaches. While some software allows transient velocity is given by the following relationship:
multiphase flow simulation using a dynamic two-fluid model,
the steady-state approach has been included in other c = C0Vm + V0 (3)
commercial tools by applying empirical and mechanistic
methods. where C0 and V0 are empirical constants which can be obtained
Comparisons between empirical and mechanistic method according to Bendiksen et. al [10].
are made in this work by running simulations in steady-state The slug frequency can be determined by the model of Hill
simulators in order to explain the differences in the calculated and Wood [13], which is based on the equilibrium film height:
slug properties. Additionally, slug sizes for operational slugging
have been analyzed by using an alternative pseudo transient  h
v  2.68 
approach to the Lagrangian slug tracking scheme.
f s = 0.275 m 10 d (4)
d
THE STEADY-STATE APPROACH
Two types of models are usually applied for hydrodynamic This model correlates the slug frequency with the diameter,
slug analysis in steady-state commercial software: mechanistic gas-liquid slip velocity and the equilibrium holdup at the
and empirical. beginning of the pipeline. The model assumes a horizontal pipe
and uses the Taitel-Dukler [14] stratified flow model to estimate
Mechanistic Model the slip velocity and the equilibrium holdup at the beginning of
Watson [11] developed a mathematical model for the the pipe.
prediction of slug properties in gas-liquid flows for horizontal
and inclined pipelines. Travelling wave solutions of the one- Empirical Method for Hydrodynamic Slug Analysis
dimensional averaged mass and momentum equations were Empirical methods develop correlations from laboratory
analyzed to obtain slug flow properties such as film holdup, and field data to predict slug length, slug frequency and slug
slug length, bubble length and average pressure gradient as a volume. The most common hydrodynamic slug correlations for
function of slug frequency. Stratified flow is tested for large diameters include SSB (Scott & Brill) and Norris
instability to small disturbances and then analyzed in the correlations.
unstable region to find if slug flow is possible. In this region a Brill et. al [15] developed a correlation based on four data
range of frequencies is possible which is predicted by the sets from Pruedhoe Bay data. The mean slug length correlated
model. with the mixture velocity and diameter of the pipe is expressed
If experimental data is not available this model can be as follows:
combined with empirical correlations for slug holdup, _ 
translational velocity and slug frequency. ln L s  = −2.663 + 5.441(ln d )0.5 + 0.059(ln vm ) (5)
 
The bubble length Lbubble is given by:  

2 Copyright © 2013 by ASME

 ∂
Norris [16] modified the Brill correlation and added data
(H G ρG ) = − 1 ∂ ( AH G ρG uG ) + ψ G + GG (7)
from one of the Prudhoe Bay experiments in 24 inches pipe ∂t A ∂z
diameter. The Norris correlation excludes the mixture-velocity ∂ 1 ∂
term. This correlation is shown as follows: (H L ρ L ) = − ( AH L ρ Lu L ) − ψ G H L
∂t A ∂z H L + H D (8)
− ψ e + ψ D + GL
 _ 
ln Ls  = −2.099 + 4.859(ln d )0.5 (6)
 
  The momentum conservation equations for the gas and
liquid phase are defined as follows:
The Brill and Norris correlations have a number of
limitations. These models do not scale up to large diameter
pipes and none of these correlations consider slug growth.
∂
∂t ∂z A ∂z
(
(H G ρG uG ) = − H G ∂p − 1 ∂ AH G ρG uG 2 )
Scott et. al. [17] modified the Norris correlation to predict 1 SG 1 S
the slug length as a function of pipe diameter. The equation is − f G ρ G uG uG − fi ρG u R u R I (9)
2 4A 2 4A
the following one: + H G ρG g cos α + ψ G u a − FD

 
ln Lsr  = −25.4144 + 28.4948(ln d )0.1 (7)
∂
∂t ∂z A ∂z
(
(H L ρ Lu L ) = − H L ∂p − 1 ∂ AH L ρ Lu L 2 )
 
  1
− f L ρL uL uL
SL 1 S
− fi ρG u R u R I
2 4A 2 4A
(10)
The resulting equations for the growth prediction are HL
+ H L ρ L g cos α + ψ G − ψ e ui
summarized as follows: HL + HD
_
∂H L
Ls = Lsr GD (8) + ψ D u D − H L D(ρ L − ρ G )g sin α
∂z

GD =  Ax + GD x 
ln x
− Ax (9) The two-fluid model OLGA includes balances for the
 t  ln xt
liquid droplets traveling within the gas phase and constitutive
equations for heat and mass transfer between the phases [18].
The Scott Correlation (SSB) is restricted to the following The constitutive equations are obtained by observing and
conditions: horizontal pipes, d > 6 inches, 1.5 < VSG < 15 m/s analyzing the phenomenon through experimental experiences,
(superficial gas velocity) and 0.3 < VSL < 1.5 m/s (superficial which are later translated into equations, direct relations and
liquid velocity). In addition, slug flow must indeed exist; additional models that provide complementary information to
therefore flow-pattern maps must be consulted. the general balances [19].
One of the modifications introduced by the OLGA model is
the reformulation of the equation system before solving and
TRANSIENT APPROACH discretizing the differential equations in order to obtain the
In a transient model, the pipeline is generally subdivided pressure relation. The model solves this pressure equation with
into a number of sections of a specific length. The pressure and the momentum conservation balance for each phase using a
the gas and liquid holdup are then calculated for every section stepwise integration. In the mass balance presented before, an
of the pipe in a particular time “t” by solving a system of equation that describes the density as a relation of temperature
equations at a every time step [18]. composition and pressure relation is included, and after further
modifications a single equation for pressure and phase flux is
Transient Two-Fluid Model developed [10].
Transient two-fluid models are based on the basic
conservation principle and treat the interface interactions at a Slug Tracking Model
detail level. The main challenge of these models is to define the Slug tracking models predict the flow regime in long
constitutive equations to describe the interaction between the pipelines for different pipe geometry and inclinations. The
phases and the environment, such as mass, momentum and heat model is based on the control of the growth and collapse of
transfer between the fluids and the wall of the conduct [19]. slugs, where each slug or wave is treated as a separated object
The general equations of the two-fluid model are described [6].
by several authors [1, 10, 18, 19], shown as mass conservation The slugs are assumed to be “found” by using a flow
equations: regime map. The position of each slug tail and front is
monitored in Lagrangian coordinates with time and the
properties are usually calculated using empirical models [1, 20].

3 Copyright © 2013 by ASME

 u u
Slug Capturing Model u S = SG − SL (14)
HG H L
The slug flow regime is predicted as a transient one-
dimensional two-fluid model. Slugs development, growth,
The slip velocity is assumed as a function of the liquid
merge or collapse depend only on the solution of the transport
holdup and the Eqn. (14) is rearranged to obtain an expression
equations for mass and momentum conservation of each phase.
for determining the superficial liquid velocity, as a function of
The basis of the two-fluid model is the formulation of two set of
the slip velocity, the liquid holdup and the mixture velocity.
equations for the momentum, mass and energy balance for each
of the phases and the interaction between the phases should be
described again by constitutive relations. The energy balance is u SL = u S H L 2 + (u M − u S )H L (15)
usually neglected assuming an isothermic pipeline [1].
Renault [21] developed a method for estimating the slug The gas velocity is calculated using the general drift flux
initiation in tracking scheme, using the slug capturing two-fluid closure relationship. It is assumed that the gas phase velocity is
model presented by Issa and Kempf [1]. This method is equal to the translation velocity of the bubbles.
considered as intermediary between the slug capturing and slug The model uses a simplified mass balance. When a mass
tracking scheme and it is able to capture directly the slug balance is performed within a computational cell and assuming
initiation process and to track the motion of every single slug in that the fluids are incompressible and the geometry of the pipe
the pipe without numerical diffusion. is constant, the mass balance can be written in terms of the
holdup and the phase superficial velocity:
A NEW PSEUDO-TRANSIENT HYDRODYNAMIC SLUG
MODEL ∂H L ∂u SL
A pseudo transient calculation model is a time dependent + =0 (16)
∂t ∂z
model that ignores some flow characteristics of transient nature.
Even though most of the hydrodynamic slug estimation models The product of the gas velocity and the liquid holdup is
introduce a mechanistic system of equations and a tracking subtracted and added to the liquid superficial velocity in Eqn.
Lagrangian frame, they perform poorly and the obtained results (16), leading to the following flux equation:
are strongly influenced by the user input data [22].
In this new model, based on the previous work made by
Danielson [23], it is assumed that the gas holdup (or void
+
(( 0 0 )
∂H L ∂ u SL − uG H L + uG ∂H L
=0
) (17)
fraction) and the liquid holdup are defined as the ratio between ∂t ∂z
the phase flow area and the effective pipe flow area. These
ratios can be also expressed in terms of the gas and liquid The proposed differential Eqn. (17) is discretized using a
velocities as follows: finite volume approximation.
The differential equation is solved using the classical
u forward Euler method, which is a time step method that
H G = SG (11)
uG considers the differential equation at every node and calculates
the value at next time step using the value previously computed
u
H L = SL (12) [24].
uL The forward Euler method implemented in the program is
as shown in Eqn. (18)
The holdup is a number that represents the fraction of each
phase in the system. This means that the addition of the liquid H Lt + ∆t = H Lt + λ (Fi −1 − Fi ) (18)
and gas holdup fractions leads to the unity. The liquid holdup
has a physical value between zero and one.
The parameter λ represents the ratio between the time step
HG + H L = 1 (13) and the spatial step.

∆t CFL
The slip velocity is defined as the difference between the λ= = (19)
gas velocity and the liquid velocity. Using the definition of the ∆x max{u0 ....u N }
liquid and gas holdup as shown in Eqn. (12) and (13), the slip
velocity difference can be written in terms of the superficial CFL is the Courant-Friedrich-Lewy factor, which must be
velocities of the phases: between zero and one for the stability of the forward Euler
method [25]. In this work, the maximum velocity is the velocity

4 Copyright © 2013 by ASME

of the gas phase uG 0 , which is a function of the mixture new holdup dependency and an updated constant parameter to
the original equation. No modifications to the velocities of the
velocity. Therefore, the λ parameter of Eqn. (18) depends of the phases expressed as a function of the fluid properties were
mixture velocity, as follows: made.
The slug length is influenced by velocity of the phases [29,
∆t CFL
λ= = = f (u M ) (20) 30]. The length of the slug is proportional to the gas velocity.
∆x uG0 An increase in the velocity of the gas phase will generate an
increment of the slug length. The superficial gas velocity was
Positivity failure can occur in first as well as in higher removed from the equation and replaced by the gas velocity.
order schemes, which can lead to an ill-posed system, blowing The gas velocity is larger than the superficial velocity and in the
up the numerical solution. The differential equation used by the developed slug model, the liquid superficial velocity and the
Danielson model [23] generates instabilities that lead to ill- liquid holdup are a function of it. Furthermore, the constant
posed problem and therefore a blowing up in the numerical parameter of the equation was increased in order to increase the
solution. A positive preserving scheme was used to limit the value of the calculated slug length. These modifications are
liquid holdup between zero and one. shown in Eqn. (22) and (23).
The method is based on a simple flux limiter, which detects

( )n
the critical numerical fluxes that lead to the instabilities and
corrects them in order to satisfy a positive preserving condition. LS u *
= 37.3776 G (22)
The method estimates a weight using a convex combination of d (uSL )1.67
the solver and performs the limitation averaging the high-order
numerical flux with a first order Lax-Friedrichs flux [25].
uG
uG * = (23)
Constitutive Relations  ρ 
Two constitutive relations are introduced in the model: the gd 1 − G 
 ρL 
parameters for the closure drift flux relation and the slip
velocity relation.
Drift Model. The drift model is based on the assumption The exponent “n” and the average liquid holdup were
that the gas velocity is almost only sensitive to mixture velocity. correlated using a parametric polynomial function, as it is
Therefore it is correct to assume that the gas velocity is constant shown in Table 1.
over the time and equal to the translational velocity of the gas
bubbles given by the closure drift flux relation. Table 1. EXPONENT “n” AS A PARAMETRIC
The parameters of the drift flux relation have been POLYNOMIAL FUNCTION
calculated using the model developed by Choi et. al. [26].
Slip Velocity. For systems with small holdup (tending to Average Liquid Holdup Correlation
zero), the liquid phase is pushed to the walls of the pipe (i.e. 4
n = −1.8295H Lave 3
+ 5.34657H Lave
annular flow regime) and the slip velocity tends to the gas
H Lave ≥ 0.378 2
velocity. Using this consideration, the slip velocity can be − 5.7034H Lave + 2.8169H Lave
approximated with the following equation:
+ 3.2832
3
n = −5531.8 H Lave 2
+ 6857.8H Lave
u S = uG − u M H L (21) 0.343 ≤ H Lave < 0.378
− 2785.5 H Lave + 375.59
Slug Length Calculation 3 2
Three pipelines with elevation profile were studied. The n = −69.91H Lave + 22.719H Lave
Other cases
first case correspond with four analysis performed by Brill et al. + 13.501H Lave − 0.7212
[15] to the Prudhoe Bay pipeline.
The second pipeline corresponds with the profile of a flow The Eqn. (23) and the exponent “n” of the Table 5 were
line system offshore of West Africa, which was analyzed by obtained using a CFL number of 0.2. With a fixed CFL number,
Burke et al. [27]. The data was acquired using a nuclear flow spatial stepping and maximum gas velocity ,which is a function
densitometer and a fast volume weight meters for a constant of the mixture velocity, there is no possible variation of the time
input of gas and liquid rate. The third pipeline is a case studied stepping.
in an in-house project, which was analyzed by using a
commercial transient simulation software. Important Features of the New Model
The data of the studied pipelines were fitted in order to The developed model estimates the slug properties for
develop a new slug length equation, from the Wang et al. [28] horizontal pipes and pipelines with an elevation profile; it takes
expression for transient analysis. The modification introduces a

5 Copyright © 2013 by ASME

into account, when requested by the user, the density variation Table 2. SUMMARY OF DATA OBTAINED IN TESTS 6, 8,
(density profile) and does not consider the drift flux closure 13 AND 14
relation as constant values, but as profile and pipe geometry
dependency. Parameter Test 6 Test 8 Test 13 Test 14
An Excel-VBA tool was developed by including the new ID
375.14 375.14 375.14 375.14
pseudo-transient hydrodynamic slug model, linked to one of the (mm.)
commercial simulation software in steady-state. The user can qo
8,491.6 9,805.4 11,469.8 11,345.3
choose the option of importing input variables from the (Sm³/D oil)
commercial software into the Excel-VBA tool, such as vapour Inlet P
4,240.3 4,309.3 4,309.3 4,357.5
and liquid density, superficial vapour and liquid velocity, (kPag)
average holdup, superficial tension, etc. Outlet P
3,826.6 3,998.9 3,826.6 3,826.6
(kPag)
Excel-VBA interface Inlet T
55.6 57.8 53.3 53.3
The corresponding Excel-VBA tool interface is shown in (°C)
Figure 1. The user can either specify or import input variables Outlet T
52.8 58.3 50.6 50.6
from the Software A. Main results of the tool include the (°C)
maximum slug length, maximum slug volume, average slug
length and average slug volume. Simulation Model Description
The production facility at Prudhoe Bay field is modeled in
two steady-state commercial simulation software by using the
mechanistic (software A) and empirical model (software B),
respectively. The elevation profile of the pipeline is shown in
Figure 2.
4

2
ELEVATION CHANGE [m]

0
Figure 1. EXCEL-VBA INTERFACE -2

-4
COMPARISON BETWEEN EMPIRICAL AND
MECHANISTIC METHOD (STEADY-STATE) -6
Existing oil and gas production facilities are modeled in -8
two commercial multiphase flow simulators. Based on these
simulation models the empirical and mechanistic methods for -10
0 1000 2000 3000 4000 5000
hydrodynamic slug analysis are compared with experimental DISTANCE FROM DRILL SITE [m]
data from the facilities.
Figure 2. ELEVATION PROFILE OF TESTED FLOW LINES
Experimental Data
The two-phase tests in Prudhoe Bay field for large diameter The fluid PVT behavior is represented by the Black Oil
flow lines published by Brill et al. [15] serve as data source for model based on the water cut, the gas/oil ratio (GOR), the gas
the comparison. A total of 29 two-phase flow tests were specific gravity and the oil gravity at stock-tank conditions.
conducted in two 3-mile-long flow lines in the Prudhoe Bay Specified boundary conditions at the source are operating
field of Alaska. Of these, 11 were for a 12-in.-diameter line and pressure, temperature and the liquid volume flow rate at stock-
18 were for a 16-in. line. 9 of the tests were in slug flow and 20 tank conditions. The selected multiphase flow correlation is
were in froth flow. Due to statistical reasons only test numbers Beggs & Brill [31]. Table 3 shows fluid properties and
6, 8, 13 and 14 are considered in this comparison. Table 2 boundary conditions.
summarizes the obtained data in the tests.

6 Copyright © 2013 by ASME

Table 3: FLUID PROPERTIES AND BOUNDARY Table 5. MEASURED AND CALCULATED MAXIMUM
CONDITIONS USED FOR THE MODELS SLUG LENGTHS

Parameter Test 6 Test 8 Test 13 Test 14 Software Software

Measured % %
Water Cut Test A1 B2
0 0 0 0 Error Error
(%) (m) (m) (m)
GOR 6 422.5 387.1 8.4 642.5 -52.1
155.4 121.7 114.7 126.4
(Sm³/Sm³) 8 398.7 318.5 20.1 666.6 -67.2
Gas specific 13 301.4 456.0 -51.3 663.2 -120.0
gravity 0.781 0.781 0.781 0.781 14 482.5 890.9 -84.6 651.7 -35.1
(-)
Operating P It can be seen from Table 5 that the maximum slug lengths
4,240.3 4,309.3 4,309.3 4,357.5
(kPag) obtained with software B are very similar among all tests. This
Operating T is due to the fact that the mean slug length, which is the basis to
55.6 57.8 53.3 53.3
(°C) calculate the maximum slug length, is only dependent of the
qL flow line diameter which is the same in all tests. Since the slug
8,491.6 9,805.4 11,469.8 11,345.3
(Sm³/d) growth mechanism included in the empirical approach in
software B predicts an elongating slug length along the flow
Simulation Model Validation line, it is not surprising to find the maximum slug length at the
The simulation models are validated with measured fluid end of the flow line. According to Table 5 the maximum slug
densities from the experiments. As can be seen in Table 4, the length calculated with the empirical approach is overestimated.
calculated oil and gas densities show good agreement with the Software A gives reasonable maximum slug lengths for the
measured data. Tests 6 and 8 while in Tests 13 and 14 the slugs are also
overestimated.
Table 4. MEASURED AND CALCULATED OIL AND GAS Both steady-state methods cannot accurately estimate the
DENSITIES AT THE FLOW LINE INLETS maximum slug length in pipelines since the relative error is
significant and variable; however, the mechanistic model is
ρmeasured ρSoftware A ρSoftware B more accurate than the empirical.
Test (kg/m³) (kg/m³)) (kg/m³))
oil gas oil gas oil gas COMPARISON BETWEEN FIELD DATA, COMMERCIAL
6 833.4 39.9 834.2 40.4 832.9 39.8 TRANSIENT SOFTWARE AND THE NEW PSEUDO
8 830.3 41.2 833.2 39.6 831.0 40.1 TRANSIENT MODEL
13 833.4 41.4 835.5 39.9 834.2 40.7 The reference data include the maximum and average
14 833.4 41.7 836.1 40.6 834.0 41.4 measured slug lengths from the Prudhoe Bay data and the
system offshore of West Africa and one case performed in-
Simulation Results and Comparison house using a transient commercial simulation software. The
The commercial simulation software used in this work can input data regarding the last two cases is summarized as
determine a “maximum slug length” which is an important follows:
factor e.g. in slug catcher sizing.
In software “A” a maximum slug length is found in each Table 6. SUMMARY OF INPUT FOR WEST AFRICA AND
flow line section at the corresponding lowest slug frequency IN-HOUSE CASE PIPELINES
(due to inclination changes, slug frequencies also vary along the West In-House
Flow Parameter
flow line). Hence, the maximum slug length in the entire flow Africa Project
line occurs in the section with the lowest possible slug Vapour density, kg/m3 21.0 23.0
frequency. Liquid density, kg/m3 889.6 845.2
In software “B” the maximum slug length is derived for Superficial Vapour Velocity, m/s 3.4 7.6
each flow line section from the corresponding mean slug length Superficial Liquid Velocity, m/s 0.4 0.4
by applying a probabilistic model. It assumes a skewed log- Surface Tension, N/m 0.02 0.02
normal distribution of the mean slug length. Pipe Diameter, m 0.18 0.39
The calculated maximum slug lengths are compared with Pipe Length, m 6073.0 15481.0
the experimental data and summarized as follows: Average Holdup, fraction 0.28 0.18
Minimum Holdup, fraction 0.07 0.11

1
The maximum slug length is found at a distance of 3,301 m in all tests.
2
The maximum slug length is found at the outlet of the flow line in all tests.

7 Copyright © 2013 by ASME

 10 slug analysis, due to the transient nature of this phenomenon.
Furthermore, the new hydrodynamic slug model is proven to be
0
a reliable alternative for hydrodynamic slug analysis due to its
ELEVATION CHANGE [m]

-10 simplicity and low cost investment to be adapted to existing

-20 tools such as Excel and the fair accuracy to determine slug
lengths and volumes.
-30
0 1000 2000 3000 4000 5000 6000
-40 CONSIDERATIONS FOR HYDRODYNAMIC SLUG
-50 ANALYSIS
Hydrodynamic slugs in pipelines are usually analyzed by
-60
using a steady-state flow assurance simulator as a first
-70
DISTANCE [m]
approximation. Two main approaches have been used and
developed in these simulators: mechanistic and empirical.
Figure 3. ELEVATION PROFILE OF WEST AFRICA Empirical correlations and mechanistic models have both
PIPELINE advantages and disadvantages to be used on the hydrodynamic
slug analysis.
The elevation profile of the in-house project can be Empirical correlations are successfully used in terms of:
considered almost horizontal. - Quick and easy formulations and calculations
The reference data is compared with the results obtained by - Fair accuracy when being used within well-defined
the new hydrodynamic slug model and summarized as follows: ranges of operating conditions
However, they have limited range of applicability to predict
Table 7. MEASURED AND CALCULATED MAXIMUM slug properties where extrapolation of flow conditions is
SLUG LENGTHS required.
Mechanistic models have the following advantages:
Average LS (m) % - Flow regimes can be successfully determined in
Test New Relative inclined pipes.
Measured - They can be extrapolated and validated for a wide
Model Error
6 422.5 448.5 -6.2 range of flow conditions.
8 398.7 386.6 -3.0 Associated disadvantages of using mechanistic models are:
13 301.4 511.1 -69.6 - Complexity of formulation and understanding
14 482.5 590.2 -22.3 - Cost of implementation in commercial software
Burke - - - As it has been demonstrated in this work, both steady-state
methods cannot accurately estimate the maximum slug length in
Table 8. MEASURED AND CALCULATED AVERAGE pipelines since the relative error is significant; nevertheless the
SLUG LENGTHS mechanistic model is more accurate than the empirical one with
lower relative errors. Therefore, it is recommended, as a first
Average LS (m) % approximation, to perform a hydrodynamic slug analysis using
Test New Relative software that include mechanistic model for the slug length
Measured calculation.
Model Error
6 156.0 141.6 9.2 It is recommended to perform a transient simulation as a
8 135.0 130.0 3.7 next step of this analysis. It has been proven that, due to the
13 135.3 169.7 -25.4 intrinsically transient nature of the hydrodynamic slug
14 183.2 187.5 -2.3 phenomenon, more accurate results regarding slug properties
Burke 98.0 98.1 -0.1 are expected. Commercial transient software have been
successfully used in this stage of the analysis, however high
Since the only available result from the in-house simulation licensing cost represents a considerable disadvantage inherent
case is the average slug volume, this is not included in the to the use of these software. For this reason, an alternative
previous tables. The average slug volume is calculated as 6.2 m3 pseudo-transient hydrodynamic slug model has been proposed
with a relative error of 1.6% compared to the value calculated in order to estimate the average and maximum slug length and
by a commercial transient simulation software. volumes in pipelines. The developed model estimates the slug
As it can be seen from Table 7 and Table 8, the average properties for horizontal pipes and pipelines with an elevation
error percentage decreases in all tests except for test 13 for the profile; it takes into account, when requested by the user, the
maximum slug length, being the transient approach more density variation (density profile) and does not consider the
accurate than the steady-state approach for the hydrodynamic drift flux closure relation as constant values, but as profile and
pipe geometry dependency.

8 Copyright © 2013 by ASME

 By using the new hydrodynamic model, the average error ρL Liquid density
percentage decreases considerably compared to the steady- ΨD Droplets mass transfer coefficient
approach results, demonstrating that the model is potentially a ΨE Entrainment gas mass transfer coefficient
reliable alternative for hydrodynamic slug analysis. ΨG Gas mass transfer coefficient
Nevertheless, further tests must be performed in order to
improve the accuracy of the slug length calculation approach by REFERENCES
modifying the constant parameter and the set of correlations to [1] Issa, R., and Kempf, M., 2003. “Simulation of Slug Flow in
estimate “n” included in Eqn. (22). Horizontal and Nearly Horizontal Pipes with two-fluid
model”. International Journal of Multiphase Flow, 29(1),
NOMENCLATURE Jan, pp. 69–95.
A Flow area [2] Ansari, M., and Shorki, V., 2011. “Numerical Modeling of
Ax 5.29 when x ≤ xt and Ax = 2.35 when x > xt Slug Flow Initiation in a Horizontal Channels using a Two-
c Slug translational velocity Fluid Model. International Journal of Heat and Fluid Flow,
CFL Courant-Friedrich-Lewy factor 32(1), Feb, pp. 145-155.
d Diameter of the pipe [3] FEESA, 2003. Hydrodynamic Slug Size in Multiphase
FD Gas / droplets drag term Flowlines. See also URL http://www.feesa.net/.
fI Friction factor at the interface [4] Fabre, J., and Line, A., 1992, “Modelling of Two-Phase
fG Friction factor in the gas phase Slug Flow”. Annual Review of Fluid Mechanics, 24, Jan,
fL Friction factor in the liquid phase pp. 21-46.
fs Slug Frequency [5] Bonizzi, M., and Issa, R., 2003, “A Model for Simulating
g Acceleration of gravity Gas Bubble Entrainment in Two-Phase Horizontal Slug
GD Dimensionless slug growth Flow”.International Journal of Multiphase Flow, 29(11),
GG Possible gas source in the system Nov, pp. 1685-1717.
GL Possible liquid source in the system [6] Carpintero, E., 2009, “Experimental Investigation of
h Film height Developing Plug and Slug Flows”. PhD thesis, Technische
HD Droplets holdup Universität München.
HG Gas holdup [7] Fan, Z., Lusseyran, F., and Hanratty, T., 1993, “Initiation of
HL Liquid holdup Slugs in Horizontal Gas-Liquid Flows”, AIChE Journal,
HLAVE Average liquid holdup 39(11), Nov, pp. 1741-1753.
ID Inner diameter [8] Sanchis, A., Johnson, G., and Jensen, A., 2011, “The
Ls Slug length Formation of Hydrodynamic Slug by Interaction of Waves
Lsr Slug length at the end of the pipeline in Gas-Liquid Two-Phase Pipe Flow”, International
n Adjustment polynomial function Journal of Multiphase Flow, 37(4), May, pp. 358 – 368.
P Pressure [9] Brill, J., and Murkherjee, H., 1999. Multiphase Flow in
qg Gas Volumetric Flow Rate Wells. Society of Petroleum Engineers. ISBN:978-1-55563-
qo Oil Volumetric Flow Rtae 080-5.
SG Gas perimeter [10] Bendiksen, K., Maines, D., Moe, R., and Nuland, S., 1991,
SI Interface perimeter “The Dynamic Two-Fluid Model OLGA: Theory and
T Temperature Application”, SPE Production Engineering, 6(2), May, pp.
ug Gas velocity 171-180.
ul Liquid velocity [11]Watson, M., 2001, “The Modeling of Slug Flow
ur Relative velocity Properties”, 10th International Conference Multiphase ’01,
us Slip velocity Jun, pp. 13-15.
usg Superficial velocity of the gas [12]Gregory, G., and Scott, D., 1969, “Correlation of Liquid
usl Superficial velocity of the liquid Slug Velocity and Frequency in Horizontal Cocurrent Gas-
um Velocity of mixture Liquid Slug Flow”. AIChE Journal, 15(6), Nov, pp. 933-
vm Velocity of mixture, ft/s 935.
x Distance from point of initiation of slug flow, ft [13]Hill, T., and Wood, D., 1994, “Slug Flow: Occurrence,
xt Point of transition from developing to long-term Consequences, and Prediction”, University of Tulsa
slug flow Centennial Petroleum Engineering Symposium, pp. 29-31.
z Distance differential [14]Taitel, Y., and Dukler, A., 1976, “A Model for Predicting
α Angle with gravity vector Flow Regime Transitions in Horizontal and Near
λ Ratio between the time step and the spatial step Horizontal Gas-Liquid Flow”. AIChE Journal, 22(1), Jan,
ρG Gas density pp. 47-55.

9 Copyright © 2013 by ASME

[15] Brill, J., , A., 1981, “Analysis of Two-Phase Flow Tests in [31]Beggs, D., and Brill, J., 1973, “A Study of Two-Phase
Large Diameter Flow Lines in Prudhoe Bay Field”. SPE Flow in Inclined Pipes”. Journal of Petroleum Technology,
Journal, 21(3), pp. 363-378. 25(3), May, pp. 607-617.
[16] Norris, L., 1982, “Correlation of Prudhoe Bay Liquid Slug
Lengths and Holdpus Including 1981 Large Diameter
Flowline Tests””. Internal Report, Exxon Production
Research Co., Houston, Texas.
[17] Scott, S., Shoham, O., and Brill, J., 1989, “Prediction of
Slug Length in Horizontal, Large-Diameter Pipes”. SPE
Production Engineering, 4(3), pp. 335-340.
[18] Taitel, Y., and Barnea, D., 1997, “Simplified Transient
Simulation of Two-Phase Flow using Quasi-Equilibrium
Momentum Balances”, International Journal of Multiphase
Flow, 23(3), Jun, pp. 493-501.
[19] Henau, V., and Raithby, G., 1995, “A Transient Two-Fluid
Model for Simulation of Slug Flow in Pipelines – ii.
Validation”, International Journal of Multiphase Flow,
21(3), Jun, pp. 351-363.
[20] Taitel, Y., and Barnea D., 1997, “Simplified Transient
Simulation of Two-Phase Flow using”, International
Journal of Multiphase Flow, 21(3), Jun, pp. 351-363.
[21] Renault, F., 2007, “A Lagrangian Slug Capturing Scheme
for Gas-Liquid Flows in Pipes”. PhD thesis, Norwegian
University of Science and Technology.
[22] Danielson, T., 2011, “A Simple Model for Hydrodynamic
Slug Flow”. Offshore Technology Conference.
[23] Danielson, T., 2012, “Transient Multiphase Flow: Past,
Present and Future with Flow Assurance Perspective”.
Energy Fuels 26(7), Jun, pp.4137-4144.
[24] Quarteroni, A., and Saleri, F., 2006, “Scientific Computing
with Matlab and Octave”. Springler.
[25]Xu, H., Adams, N., and Shiu, C., 2013, “Positive-
preserving Method for High-Order Conservative Schemes
Solving Compressible Euler Equations”. Journal of
Computational Physics, 242, Jun, pp.169-180.
[26]Choi, J., Pereyra, E., Sarica, Park, C., and Kang, J., 2012,
“An Efficient Drift-Flux Closure Relationship to Estimate
Liquid Holdups of Gas-Liquid Two-Phase Flow in Pipes”.
Energies, 5(12), Dec, pp.5294-5306.
[27]Burke, N., and Kashou, S., 1996, “Slug-Sizing/Slug-
Volume Prediction: State of Art Review and Simulation”.
SPE Production and Facilities, 11(3), pp.166-172.
[28]Wang, X., Guo, L, and Zhang, X., 2006, “Slug-
Sizing/Slug-Volume Prediction: State of Art Review and
Simulation”. Chinese Journal of Chemical Engineering,
14(5), Oct, pp.626-633.
[29]Cai, J., 1999, “Slug Frequency and Length Inclined Large
Diameter Multiphase Pipeline”. Multiphase Flow and Heat
Transfer, Aug, pp.22-24.
[30]Wang, X., Guo, L, and Zhang, X., 2007, “An Experimental
Study of the Statistical Parameters of Gas-Liquid Two-
Phase Slug Flow in Horizontal Pipeline”. International
Journal of Heat and Mass Transfer, 50(11-12), Jun, pp.
2439-2443.

10 Copyright © 2013 by ASME

View publication stats